Use truncated and terminated return values correctly

[clemens-fricke]

Is your feature request related to a problem? Please describe.
When this package was first written, stable_baselines3 did not differentiate between different reasons for terminating an episode. But a few versions ago this changed. There are now two different episode termination variables:
terminated and truncated

I at some point saw this change and did a quick and dirty fix (I think) in the framework, but did not understand what the difference was back then. Now I just came by this part of the sb3 documentation that goes a little bit into the difference link to the docu part.

This needs to be implemented correctly so that we can give the agents the best possible information to train from.

Describe the solution you'd like
Adapt the correct variable to the end state if the episode is terminated.